Fit a Model on a Single ROI

The primary dispatch point for per-ROI analysis in the new architecture. Each model type implements this method to encapsulate its complete analysis pipeline (including any internal cross-validation).

Computes mediation paths a, b, c' and the indirect effect in RDM space for a single ROI/searchlight.

Computes representational connectivity metrics for a single ROI/searchlight.

Usage

fit_roi(model, roi_data, context, ...)

# S3 method for class 'manova_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'mvpa_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'item_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'rsa_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'vector_rsa_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'feature_rsa_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'naive_xdec_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'era_partition_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'repmap_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'repmed_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'repnet_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'subspace_alignment_model'
fit_roi(model, roi_data, context, ...)

# S3 method for class 'feature_rsa_da_model'
fit_roi(model, roi_data, context, ...)

Arguments

model

The model specification object.

roi_data

A list with components:

train_data

Numeric matrix (observations x features)

test_data

Numeric matrix or NULL

indices

Integer vector of voxel/vertex indices

train_roi

The raw ROI object (for models needing spatial info)

test_roi

The raw test ROI object or NULL

context

A list with components:

design

The design object

cv_spec

Cross-validation specification or NULL

id

ROI identifier (center voxel ID or region number)

center_global_id

Global index of center voxel or NA

...

Additional model-specific arguments.

Value

A roi_result object.

Details

Models that implement fit_roi are automatically preferred by process_roi.default over the legacy dispatch chain (internal_crossval / external_crossval / train_model).

Architecture TODO

fit_roi is currently scoped to ROI/searchlight iteration. Whole-brain global analysis uses run_global via a separate cv_run_global -> train_model path. A future refactor may unify these flows under a single fit contract.

See also

roi_result, output_schema, process_roi

Examples

# \donttest{
  # fit_roi is typically called internally by process_roi.default.
  # To implement for a new model class:
  # fit_roi.my_model <- function(model, roi_data, context, ...) {
  #   metrics <- c(accuracy = 0.85)
  #   roi_result(metrics, roi_data$indices, context$id)
  # }
# }